Neurocognitive Adaptation and Mental Health Vulnerability Following Maltreatment: The Role of Social Functioning

Evidence of Maladaptive Social Functioning Following Childhood Maltreatment

Social functioning—defined broadly as an individual’s ability to perform and fulfil normal social roles (Hirschfeld et al., 2000)—is arguably both an important mediator in the development of psychopathology following childhood maltreatment and an important developmental outcome in its own right. How maltreatment impacts social functioning has been investigated in a broad range of cross-sectional and longitudinal studies. One line of research has investigated how social competence is perceived by others. In a large national cohort of kindergarten children (N = 69,116), associations were observed between exposure to any one type maltreatment and an individual’s odds of being ranked in the bottom 10% of the cohort for teacher-rated social competency (OR = 2.7; CI [2.3, 3.1]), even after adjusting for a range of potential confounds such as age, sex, socioeconomic status, and parental schizophrenia (Matheson et al., 2017). This effect was even more pronounced for those who had experienced two or more subtypes of maltreatment (OR = 5; CI [4, 6.4]). In a study of young adults (N = 435), harsher peer first impressions were found to be formed of those who had witnessed domestic violence in childhood (King, 2016).

Compelling evidence that the impact of childhood maltreatment (and adversity more generally) has an enduring effect on social competence comes from recent longitudinal studies that have looked at social functioning across several periods with respect to the same individuals. For example, Raby and colleagues used data from the Minnesota Longitudinal Study of Risk and Adaptation (MLSRA; N = 267) to evaluate whether abuse and neglect during the first 5 years of life had long-term consequences for social competence. Teacher reported competence with peers was measured up to age 16, followed by competence at forming and maintaining high-quality romantic relationships in adulthood (Raby et al., 2018). Maltreatment was found to be directly associated with lower levels of social competence at each developmental stage (Grades 1–6 and ages 16, 23, and 32), which remained the case even after accounting for the stability of social adjustment across development. Similarly, Flynn, Cicchetti, and Rogosch (2014) employed a three-wave longitudinal design (N = 635) to test the effects of childhood maltreatment on low self-worth, low-quality relationships, and both internalizing and externalizing symptoms across early-mid and mid-late adolescence. Not only did this design allow researchers to study whether vulnerabilities associated with maltreatment in early-mid adolescence remained stable across development, they were also able to test for significant cross-lagged effects whereby a parameter in the second wave shows independent associations with the remaining parameters of interest in the third wave (Kearney, 2017). Again, difficulties in these domains endured across time points, with independent associations observed between low maternal relationship quality in Wave 2 and both internalizing symptoms and self-esteem in Wave 3.

A second line of research has explored functioning within adult romantic relationships. Drawing upon a subsample of participants from the MLSRA (N = 179), and using semistructured interviews and videotaped interactions between study participants and their partners, Labella and colleagues (2018) found that childhood abuse and neglect predicted poorer romantic competence. Kapeleris and Paivio (2011) also investigated romantic competence in a sample of 187 undergraduates. Here, participants with a more severe history of emotional abuse and neglect displayed relative difficulties identifying and expressing emotions (characteristic of alexithymia) and emotional dysregulation and inhibition (characteristic of maladaptive emotional processing), with these emotional competence measures fully mediating the association between emotional abuse and neglect and adult attachment styles in romantic relationships.

Unsurprisingly, these findings indicating poorer social competence are complemented by evidence that maltreatment experience is also associated with increased likelihood of rejection and victimization. In one study, Kim and Cicchetti (2010) asked 421 low socioeconomic status (SES) children attending a summer camp to nominate which peer they liked best and least overall. In line with several earlier results (Bolger, Patterson, & Kupersmidt, 1998; Dodge, Pettit, & Bates, 1994), emotional maltreatment, physical abuse, and sexual abuse were all predictive of higher peer rejection (being rated as “least liked”). Moreover, there is now convincing data derived from a longitudinal designs that childhood maltreatment is causally associated with an increased risk of being bullied (“bullying victimization”; Benedini, Fagan, & Gibson, 2016). Analogous patterns of rejection and revictimization have also been reported in adulthood. In a prospective study of adults (N = 892), Widom, Czaja, and Dutton (2014) found that maltreatment history was associated with an increased likelihood of both experiencing and perpetrating intimate partner violence; a result that has since been repeated in a study of over 30,000 adults who provided retrospective reports of maltreatment history (Afifi, Mota, Sareen, & MacMillan, 2017).

In addition to evidence of poorer social competence and increased risk of rejection and victimization, research has also suggested that individuals who have experienced childhood maltreatment benefit from weaker social support. Using their prospective sample, Sperry and Widom (2013) found a negative association between documented maltreatment history and self-reported social support in adulthood, 30 years later. This is significant in the context of mental health vulnerability as social support is one factor that is thought to buffer against stressful life events that potentiate risk of developing psychiatric disorders (Harkness, Lumley, & Truss, 2008; Uhrlass & Gibb, 2007). Indeed, reduced social support has been directly associated with an increased risk of anxiety and depression (Sperry & Widom, 2013; van Harmelen, 2016). Prospective cohort studies incorporating several waves of longitudinal data from childhood through to early adulthood have also found that differences in perceived social isolation partially mediate the observed associations between maltreatment and mental health symptoms in later life (Alto, Handley, Rogosch, Cicchetti, & Toth, 2018; Sheikh, 2017).

While there is a general consensus regarding the significant impact of maltreatment on social functioning and subsequent mental health outcomes, it is somewhat surprising how little we know about the neurocognitive mechanisms that may underpin this association. Yet a growing body of work now demonstrates that maltreatment is associated with altered functioning across a range of neurocognitive domains likely to be directly implicated in key aspects of social functioning.

The Theory of Latent Vulnerability: Neurocognitive Adaptation Within the Context of Social Functioning and Mental Health

Neuroimaging studies have provided evidence that maltreatment in childhood results in measurable alterations to a number of neurocognitive systems (McCrory, Gerin, & Viding, 2017; Teicher, Samson, Anderson, & Ohashi, 2016). However, how these alterations impact social functioning in ways that may increase risk of future mental health problems has been accorded little explicit focus. We have postulated in our theory of latent vulnerability that such alterations can be viewed as representing in part a functional and adaptive response to early neglectful and/or abusive environments (McCrory & Viding, 2015). However, such adaptations are equally thought to incur a longer term cost, as they may mean that the individual is poorly optimized to negotiate the demands of other, more normative environments. Here, we consider how these costs may have a particular impact on social functioning and social decision-making. While it is of course the case that patterns of adaptation arise at multiple levels (see Cicchetti, 2016), we focus on neurocognitive functioning as the level of investigation most likely to have immediate translational relevance.

We have previously suggested that these neurocognitive changes can confer latent vulnerability either directly or indirectly (McCrory et al., 2017). Direct effects can be understood as the way in which maltreatment-related neurocognitive changes alter how an individual perceives, processes, and responds to the social world around them. For example, altered emotion regulation may have a direct effect on psychological functioning, increasing the degree to which new stressors or indeed everyday challenges burden and tax an individual (Tottenham & Gabard-Durnam, 2017); this can be understood as an increase in “stress susceptibility.” Equally, neurocognitive changes may alter how an individual influences their own social experience. Direct effects here capture how individuals may act in ways that precipitate the likelihood of stressor events occurring; this can be considered a form of “stress generation.” By contrast, indirect effects refer to how maltreatment-related neurocognitive changes influence the way that an individual cumulatively shapes their own social ecology over time. As we have already noted, childhood maltreatment history is associated with an increased likelihood of reduced social support in adulthood, even decades after abuse is experienced (Sperry & Widom, 2013). Such reduced social support may make individuals more vulnerable to stressors when they occur: This is another example of increased stress susceptibility. Equally, there may be an increase in stress generation through a greater likelihood of affiliation with individuals who are violent or abusive.

While the extant neuroimaging research has documented alterations associated with maltreatment across a range of neurocognitive systems that are likely to be crucially involved in social functioning, the precise ways in which these alterations increase latent vulnerability are unclear. A mechanistic understanding of this relationship is essential in order to inform therapeutic strategies that could target distinct cognitive processes that would shift a trajectory of maladaptive social transactions across the life course. As shown below, the paradigms that have been used to date typically evoke single responses that do not involve dyadic or iterative social interactions. This makes it difficult to directly test how altered neurocognitive functioning may compromise social competence and social decision-making.

Functional Magnetic Resonance Imaging (fMRI) Studies of Childhood Maltreatment

In light of our interest in underlying neurocognitive mechanisms, we primarily focus on functional magnetic imaging (fMRI) studies of children and adolescents. As such, a systematic consideration of studies of brain structure and those using animal models is beyond the scope of this review. Interested readers are referred to other excellent reviews of these topics (see Teicher & Samson, 2016 and McLaughlin et al., 2014, respectively). Given the policy and clinical focus on the specific subgroup of children who are exposed to childhood maltreatment (including physical, sexual, and emotional abuse and neglect, as well as exposure to intimate partner violence), we prioritize studies that have investigated these populations. Given the small number of extant functional imaging studies in the field of childhood maltreatment, we also include studies of children who have experienced severe neglect as a result of early institutionalization. Other forms of childhood adversity or stress (e.g., death of a close family member, poverty, witnessing community violence) are also important for mental health functioning and neurocognitive development but are beyond the scope of the current review. To date, extant studies have demonstrated altered functioning in three neurocognitive domains implicated in social processing: threat processing, reward processing, and emotion regulation. We first review the evidence of altered functioning in these domains among individuals exposed to maltreatment (or institutionalization) in childhood and adolescence and then turn to consider how these alterations may be relevant to initiating and sustaining impaired social functioning in subsequent normative environments.

Discussion

We have considered how specific neurocognitive domains appear to be calibrated by early exposure to maltreatment and have speculated on what this may mean for social functioning as both a mediating and moderating factor in the later expression of mental illness. Although research pertaining to human brain structure and research in animals was not reviewed here in depth owing to space constraints, findings from these related areas are likely to also have important implications for our understanding of impaired social functioning following early life maltreatment (Hanson et al., 2010; McLaughlin et al., 2014; Teicher & Samson, 2016). Moreover, among the functional studies that were reviewed, there are several limitations that characterize our current understanding.

First, we need to better understand the ways in which alterations among a range of discrete neurocognitive systems may work in concert to potentiate socially deleterious functioning. For example, recent theoretical accounts of emotion regulation (Etkin et al., 2015) have conceived of regulatory actions as the final output of a complex set of perception and valuation processes involving judgments of whether the emotion experienced best serves an individual’s current goals, which themselves are a likely function of momentary affect (Huys & Renz, 2017). How a stimulus is perceived (e.g., as more or less threatening) and whether the innate and learnt values of all possible regulatory actions are successfully updated following experience (i.e., value-based learning) will exert significant influence over how emotions are finally regulated. The example of prosocial behavior serves to illustrate this point. While a prosocial behavior such as sharing has been shown to demand significant regulatory activity in the dlPFC (Ruff, Ugazio, & Fehr, 2013; Steinbeis, 2018), heightened threat reactivity may restrain this capacity. For example, incidental aversive threats operating in the background of a social exchange paradigm (i.e., threats that are unrelated to the incentive structure of the interaction) have been shown to lead to the suppression of neural activity in the dlPFC and a reduced likelihood of cooperation (Engelmann, Meyer, Ruff, & Fehr, 2018).

Second, it is important to emphasize that extant research has focused only on a limited set of neurocognitive mechanisms; it is likely that adaptations in other cognitive domains with more explicit social functions will be important in contributing to an increased risk of psychopathology. For example, psychological research over recent decades has shown a meaningful association between maltreatment and poorer performance on various indices of social emotional understanding (N. Luke & Banerjee, 2013). Impaired social understanding in turn is thought to result in a diminished ability to respond appropriately when others express their emotions, increasing the risk social exclusion (Jaffee, 2017). However, neuroimaging studies have yet to systematically explore the neurocognitive processes implicated in social understanding.

Third, it is important to acknowledge that the extant functional neuroimaging literature has yet to shed clear light on the potential differential impact of different forms of maltreatment experience. In light of the fact that polyvictimization typically characterizes the experience of individuals exposed to maltreatment (Berzenski & Yates, 2011; McLaughlin, Gureje, Kawakami, Koenen, & Kessler, 2018), it is very difficult (and often impossible) in experimental studies with small sample sizes to reliably establish whether a particular subtype of maltreatment has a unique effect on functional brain activation (Button et al., 2013; Yarkoni, 2009). In the wider literature, we know that there is limited evidence for specific maltreatment types being associated with specific outcomes: Different forms of maltreatment appear to have largely equivalent psychiatric and behavioral effects, including anxiety depression and conduct problems (Cecil, Viding, Fearon, Glaser, & McCrory, 2017; Vachon, Krueger, Rogosch, & Cicchetti, 2015). This does not necessarily imply, however, that common underlying neurocognitive mechanisms instantiate vulnerability. A recent systematic review by Cassiers and colleagues (2018) including both functional and structural studies provides some evidence for long-term differential effects of maltreatment subtypes on the human brain. One broad distinction that has been made is in relation to those forms of abuse associated with direct harm (physical, emotional, and sexual abuse) and those associated with deprivation-related experiences (such as neglect). Sheridan and McLaughlin, in an important theoretical contribution, have proposed that the former is likely associated with alterations in neural circuits that underlie emotion and emotional learning, while the latter is associated with broader neurocognitive consequences as a result of impoverished cognitive and sensory stimulation throughout development (McLaughlin et al., 2014; McLaughlin, Sheridan, & Nelson, 2016; Sheridan & McLaughlin, 2014). One recent study of young adults provides some support for this view (Puetz et al., in press). Neural reactivity to facial cues signaling threat was compared across individuals retrospectively reporting experience of active abuse versus a propensity score matched group reporting only experience of neglect (N = 87). In line with McLaughlin and Sheridan’s proposal, it was found that while childhood abuse was associated with heightened localized threat reactivity in the ventral amygdala, experiences of neglect were associated with heightened reactivity in a distributed cortical fronto-parietal network supporting complex social and cognitive processing as well as in the dorsal amygdala (Puetz et al., in press). Among the studies meeting our selection criteria for review, only two involved a direct comparison (either in primary or secondary analysis) of neural activations for exposed and unexposed individuals to distinct maltreatment subtypes controlling for the effects of other subtypes. Lee and colleagues (2015) isolated the effects of emotional abuse by excluding all participants reporting a history of other forms abuse, while Hanson and colleagues (2015) covaried the influence of other forms of abuse in order to investigate the potential unique effect of emotional neglect. Looking forward, rather than focusing on the particular form of maltreatment experience in isolation, it may be more relevant to consider timing and duration alongside. Functional imaging studies, however, have not typically measured onset and duration of maltreatment experience, despite evidence of distinct region-specific structural changes associated with maltreatment exposures at different development stages (Teicher & Samson, 2016). Moreover, a recent community-based study has highlighted the importance of going beyond a simple focus on subtype. In a latent class analysis approach with a sample of 674 low-income urban children, Warmingham and colleagues (2018) found that those who had experienced maltreatment fell into one of three groups: those who had experienced chronic, multi subtype maltreatment (57%), those who had experienced neglect only in a single developmental period (31%), and those characterized by a single subtype of maltreatment occurring in a single developmental period (12%). This is consistent with the view that a focus on a single maltreatment subtype (other than perhaps neglect) is unlikely to be particularly meaningful in relation to children’s experiences in the real world.

Finally, it is important to note that we have not considered the impact of genetic factors related to alterations in neural response in the context of maltreatment. None of the studies we have described incorporated molecular genetic data into statistical models when estimating associations between maltreatment and task-induced neural activity. Such imaging genetics approaches can support more accurate estimation by controlling for genetically determined traits that moderate associations between maltreatment and the independent variables of interest, providing greater precision when attempting to isolate key neurobiological pathways involved in both disease risk and resilience (Munoz, Hyde, & Hariri, 2009). To our knowledge, only one functional study of children and adolescents with maltreatment experience has adopted such a design. In a sample of psychiatrically healthy adolescents (N = 139), White et al. (2012) observed that degree of emotional neglect was predictive of higher amygdala reactivity during a standard emotional face-processing task, although only among those adolescents with a common variant of the FKBP5 gene linked to HPA axis dysfunction. It must be emphasized that across the wider candidate gene literature, decades of genetic case–control association studies (with and without maltreatment exposure included as a moderating environmental factor) have now yielded many publications with very few consistent replications (Arango, 2017; McCrory, De Brito, & Viding, 2011). For example, in a recent replication study of 37 adult imaging genetics studies of threat-related amygdala activity in response to the same emotional face-processing task, Avinun, Nevo, Knodt, Elliott, and Hariri (2018) found that only 3 of the 37 genetic variants studied (i.e., single-nucleotide polymorphisms, frequently called SNPs—pronounced “snips”) replicated in their sample of young adults (N = 1,117). In light of these poor rates of replication, findings derived exclusively from small experimental samples should be treated with some caution, given that the effects of single common genetic variants on intermediate neural phenotypes will likely be small.

Looking Forward: Advancing Our Understanding of Maltreatment, Social Functioning, and Resilience

Developmental neuroscientists have investigated neurocognitive alterations following maltreatment exposure that are believed to be important in understanding long-term mental health vulnerability. This research has been particularly important in highlighting that functional neural differences often develop prior to the onset of symptomatology following exposure to maltreatment. These alterations can therefore be considered “latent” in nature and, as such, give impetus to a preventative approach where help is offered before mental health problems emerge. However, across mental health research (including research into the neurocognitive effects of early maltreatment), functional neuroimaging procedures have yet to have a major translational impact on clinical practice. One explanation for this limited translational success is that to date, most neuroimaging investigations have been essentially descriptive in nature, identifying disorder-relevant circuits. Yet on their own, such neurocognitive evidence fails to provide a mechanistic account of either circuit function or complex behavior (Stephan, Iglesias, Heinzle, & Diaconescu, 2015).

Understanding how neurocognitive alterations impact social functioning, and ultimately mental health outcomes, will require us to precisely define and measure the component processes underlying social behavior. Moreover, we will need to understand these processes in the context of the dynamic interaction between multiple neurocognitive systems. For experimental researchers, developing this understanding will involve two steps. First, formalizing testable mathematical models that specify the interacting components that are hypothesized to generate each individual decision over the course of a task. Second, developing sufficiently naturalistic paradigms able to evoke affective responses typical of those elicited during real-life interactions (Moutoussis, Eldar, & Dolan, 2017). The latter may be achieved through the use of social exchange paradigms adapted from behavioral game theory (Sharp, Monterosso, & Montague, 2012), while the former reflects recent moves within psychiatry to understand the specific computations performed by the brain (Hauser, Will, Dubois, & Dolan, 2018) and how these computations may be nonoptimal in ways relevant to psychiatric functioning (Moutoussis, Story, & Dolan, 2015).

The distinct advantage of formalizing computational models in neuroscientific research is that they can help us to test verbal theories of how neurocognitive adaptations shape behavior (van den Bos et al., 2017). Here, for instance, we have speculated that hypersensitive threat processing, instantiated by increased amygdala reactivity, may impair social functioning via (i) an individual’s overattribution of threat signals, leading to (ii) more conflictual social interactions. Latent variables such as an individual’s perception of threatening stimuli or their degree of hostile attribution bias often cannot be measured directly but can instead be extracted from a computational model estimated from participants’ behavior over a task, again highlighting the need for experimental paradigms able to capture at least some of the computational richness of interpersonal exchange (King-Casas & Chiu, 2012; Krakauer, Ghazanfar, Gomez-Marin, MacIver, & Poeppel, 2017; Montague, 2018). These variables (known as model parameters) can, in turn, be regressed against measures of task-based neural activity to determine where such latent variables are computed in the brain (Charpentier & O’Doherty, 2018). At present, the majority of the computational treatments of psychiatric symptomology pertain to models of aberrant learning and decision-making (Stephan & Mathys, 2014). For example, a learning model derived from a social feedback task was recently able to account for variability in the degree to which participants updated their self-esteem following positive and negative social approval ratings, with vulnerability to feedback from others found to be correlated with insula-vmPFC connectivity during self-esteem updates (Will, Rutledge, Moutoussis, & Dolan, 2017). With researchers already combining therapies such as CBT with neuroimaging in order to test their efficacy using neuroanatomical markers of change (e.g., Mason, Peters, Williams, & Kumari, 2017), computational modeling of social learning and decision-making after childhood maltreatment exposure may provide important mechanistic insights about specific neural and behavioral markers of vulnerability. However, whether computational modeling will have a tangible impact on approaches to prevention and intervention remains to be seen. (Montague, 2018; Stephan & Mathys, 2014).

In addition to pursuing greater mechanistic clarity through computational modeling, collecting measures of social functioning in the “real world” alongside experimental data will be an important step in addressing the predictive validity of task-based neural and behavioral measures. With the noted exception of Tottenham and colleagues (2011), extant research has not systematically reported measures of social functioning outside of the scanner that can be related to task-based neuroimaging data. It will be important for future studies to include both subjective measures pertaining to social processes (e.g., relationship satisfaction or perceived social support) and more objective assessments such as third-party reports or the relative connectedness of individuals within defined social networks (Luke & Harris, 2007). Integrating these measures into neuroimaging research, as well as computational models, will be necessary to develop mechanistic accounts of how alterations in neurocognitive functioning relate to resilient outcomes following maltreatment experience. Social inputs at multiple levels of the child’s ecology (i.e., parent, peer, school, and community support) have been shown to be pivotal in the probabilistic emergence of resilient outcomes longitudinally via the fostering of social skills (Cicchetti, 2016; Oshri, Topple, & Carlson, 2017). Interventions targeting social functioning and social skills in particular have been shown to hold particular promise (Cicchetti, 2016; Oshri et al., 2017). One possibility is that such interventions help children “recalibrate” the functioning of distinct neurocognitive domains that have been shaped by aberrant social influences from carers earlier in life. We suggest that combining computational approaches, functional neuroimaging, and the longitudinal study of multidimensional measures of children’s interpersonal functioning has the potential to inform therapeutic strategies that could more precisely target distinct cognitive processes. Such targeting of specific cognitive processes increases the likelihood that we can shift patterns of maladaptive social transactions in ways that could promote more resilient outcomes over time.

Conclusions

We suggest that one key way that neurocognitive alterations following childhood maltreatment exposure impacts mental health risk is via social functioning. Specifically, such alterations may compromise the ability of an individual to negotiate stress (increased stress susceptibility) and elevate the probability of future stressful events occurring (increased stress generation). These influences on an individual’s social experience may be understood as arising in the form of either a direct effect (i.e., pertaining to online immediate social experience) or an indirect effect (i.e., pertaining to how an individual cumulatively shapes their social ecology over time). We have highlighted key findings that demonstrate impairments in social functioning and weaker social support in individuals who have experienced maltreatment in childhood. We have also reviewed the extant neuroimaging evidence in children and adolescents in three neurocognitive domains implicated in social functioning: threat processing, reward processing, and emotion regulation. We propose that altered functioning of these systems following adaptation to early adverse environments may compromise social functioning in more normative environments in the longer term. However, we are aware that other systems may be implicated in instantiating this vulnerability. Recently, for example, we have reported that adolescents who have experienced maltreatment present with significant differences in autobiographical memory functioning including a pattern of overgeneral recall that may negatively impact social functioning by making it more difficult to draw on past experiences when negotiating social situations (McCrory et al., 2017). Understanding how alterations in neurocognitive systems impact social functioning represents, in our view, a crucial step in explaining the link between early adverse experience and later psychiatric vulnerability. Such a proposal is informed by the central role that effective social functioning plays in sustaining good mental health—both in helping us successfully negotiate stress and life challenges but also in cultivating and sustaining a network of supportive relationships that help mitigate the impact of stressors when they arise.

Currently, how altered functioning of neurocognitive systems following maltreatment influence social decision-making and functioning in differing contexts and across several developmental stages is poorly understood. The development of a more precise mechanistic understanding of impaired social functioning will have important consequences for both therapeutic and preventative strategies. If we better understand how mental health problems unfold over time through the cumulative impact on social functioning, we will be much better placed to develop novel strategies to support young people who have experienced maltreatment, increasing the likelihood of a positive outcome. Preliminary evidence in the field is promising. For example, DePrince, Chu, Labus, Shirk, and Potter (2015) found that a 12-session group intervention with adolescents with histories of maltreatment, which included teaching mindfulness, problem-solving, and accurate threat detection, led to a nearly 5-fold decrease in reports of sexual victimization over a subsequent 6-month period. The active ingredients of this approach and whether they pertain to social functioning remain unclear. We believe that computational models of maladaptive social behavior, developed through the use of naturalistic social exchange paradigms, alongside more systematic measurement of social functioning across domains, have the potential to play an important role in shedding light on which aspects of our interventions drive change, as well helping us better understand and predict individual response variation.